CDCS Digital Research Prize: Best Novel Use of a Digital Method

Opening the Well ('Fosgladh an Tobair' in Gaelic) represents a novel and
coherent use of digital methods in Gaelic studies, heritage research and
language technology. Developed within the University of Edinburgh’s ÈIST
programme (eist.ed.ac.uk), the project goes well beyond digitising archival
material. It integrates automatic speech recognition (ASR) directly into a
public-facing community transcription platform, using machine-generated
first-pass transcripts to support volunteers in transcribing Gaelic oral
recordings from major heritage collections. This is a transformative
methodological intervention: Gaelic transcription has traditionally been slow,
specialist-led and difficult to scale. By embedding ASR within a structured
workflow -- including expert vetting, republication and model retraining --
Opening the Well creates a virtuous cycle in which human expertise and
machine learning reinforce one another.
The approach is particularly well suited to this linguistic context. Gaelic is an
under-resourced language in which data sparsity is a major barrier to
producing robust digital tools. This challenge is explicitly recognised in the
Scottish Government’s CivTech Challenge 11.2 on Gaelic data sparsity (see
link below). Rather than treating the issue solely as a technical problem,
Opening the Well addresses it through a humanities-centred design:
volunteers transcribe culturally significant recordings requested through Tobar
an Dualchais / Kist o Riches; the community-produced transcription are
quality controlled; and the resulting material simultaneously enriches public
heritage access and improves Gaelic technology. As such, the project
combines archival practice, community participation and speech technology in
a way that is both rigorous and scalable. The open-source platform developed
for the project is inspiring similar initiatives, including adaptation for Irish-
language heritage materials (in development at Dublin City University).